#!/usr/bin/env bash
set -euo pipefail

unset GGML_CUDA_FORCE_MMQ

echo ">> GGML_CUDA_FORCE_MMQ: (unset)"
echo ">> GGML_CUDA_FORCE_CUBLAS: ${GGML_CUDA_FORCE_CUBLAS:-<not set>}"
echo ">> LLAMA_LOG_LEVEL: ${LLAMA_LOG_LEVEL:-<not set>}"

exec python3 -m llama_cpp.server "$@"